[linux-audio-user] xruns vs. disk activity

[rob fell]

Hi Mark

> Normally there isn't much you can do about this, other than put in some
> other ATA disk controller or go SCSI, which I wouldn't recommend just
> yet. Fortunately I don't think any of that is really necessary just yet.
> These are good experiments though. On the NIC side, just unpluging the
> cable would normally be enough to test.

well - i didn't unplug the cable, but i did an ifconfig eth0 down, which AFAIK 
closes the interface entirely. i'm behind a firewall too, so i shouldn't be 
exposed to the usual  babble of ill-config'd M$ machines.

i'm willing to be disabused by someone knowledgeable here.

> This is where I think we really start getting into PC chipset design.
> (something I used to do for AMD.) Without belaboring the details, a good
> card is capable of getting data to and from memory without much
> processor interaction and very efficiently on the PCI bus. If it does
> that, then what's left is basically how it tells the processor what's
> going on, which is the interrupt service routine's job. However, even a
> great sound card can be hindered by other parts of the PC (like other
> PCI cards) usign the bus badly.
>
> You don't happen to be using a PCI VGA adapter, do you? I Cannot
> remember at this point.

i wish it were so easy - i have AGP.

> > > the most success so far with the testing seems to be at 1024 - though
> > > large file copies causes massive xruns at any rate (BTW: how large
> > > would a 'large' file be?).
>
> This sounds like an interrupt problem, or else a bad set of hard disk
> device drivers. If you sound card want to interrupt, it should get
> serviced even if the hard disk was on the bus. Possibly the DMA
> controller in this ATA chipset isn't very efficient....

this is sounding expensive...

surely others in the LL audio world are using the "IDE interface: VIA 
Technologies, Inc. VT82C586/B/686A/B PIPC Bus Master IDE (rev 06)" - has 
anyone anything good/bad to say about this (BTW: i did enable all the 'fix 
borken VIA' stuff in the kernel options).

> I think I'd maybe try running Benno's disk latency test first. IF it
> says you have long latency issues in your disk subsystem, then a newer
> sound card isn't going to do you much good.

it segfaults for me. it inspired me to try to use the jackd output as a guide 
for tuning the machine since i realised i wouldn't get far without some sort 
of measurement/instrumentation.

so my current alternatives are;
1. post my results
2. learn gdb and try to get latency test to work
3. buy a new snd card. pray.
4. buy a PCI IDE card. waste a lot of time. pray.
5. buy a new mobo whose chipset is recommended.

i'll post again when i've organised my test results. hopefully someone smarter 
than i can make some sense of it all (well audio-wise anyhow) and point me in 
the right direction.....

thanks all for your assistance thus far
R